The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The diagnostic scripts scripts/check_ws_smoke.py and scripts/ws_smoke_test.py represent a potential indirect injection surface.
Ingestion points: The --url command-line argument in both scripts triggers connections to external servers.
Boundary markers: The scripts lack explicit instructions to the agent to ignore potentially malicious content in server responses.
Capability inventory: The scripts use socket.create_connection and the websocket-client library to establish network connections.
Sanitization: Input is parsed using the standard urllib.parse library. Given the purpose of these scripts is technical diagnostics and they do not execute server-returned data as code, this finding is considered safe for the intended use case.
[EXTERNAL_DOWNLOADS]: The data/sources.json file references technical documentation from well-known and trusted entities, including Mozilla (MDN), Cloudflare, Apple (WebKit), and Redis. These are legitimate informational resources.
[SAFE]: The skill's documentation and code perform standard real-time systems implementation and testing. No evidence of prompt injection, obfuscation, data exfiltration, or persistence mechanisms was found. The logic aligns with best practices for WebSocket and CRDT-based development.

software-realtime